Partitioning of Serpin-Proteinase Reactions between Stable Inhibition and Substrate Cleavage Is Regulated by the Rate of Serpin Reactive Center Loop Insertion into b -Sheet A*

The serpin family of serine proteinase inhibitors is a mechanistically unique class of naturally occurring proteinase inhibitors that trap target enzymes as stable covalent acyl-enzyme complexes. This mechanism appears to require both cleavage of the serpin reactive center loop (RCL) by the proteinase and a significant conformational change in the serpin structure involving rapid insertion of the RCL into the center of an existing b -sheet, serpin b -sheet A. The present study demon-strates that partitioning between inhibitor and substrate modes of reaction can be altered by varying either the rates of RCL insertion or deacylation using a library of serpin RCL mutants substituted in the critical P 14 hinge residue and three different proteinases. We further correlate the changes in partitioning with the ac-tual rates of RCL insertion for several of the variants upon reaction with the different proteinases as determined by fluorescence spectroscopy of specific RCL-la-beled inhibitor mutants. These data demonstrate that the serpin mechanism follows a branched pathway, and that the formation of a stable inhibited complex is dependent upon both the rate of the RCL conformational change and the rate of enzyme deacylation. The serpins are as of PAI-1 mutant. Kinetic Analysis of the P 9 -NBD-PAI-1 Variants— For fast reactions stopped flow fluorometry was performed on PAI-1 NBD derivatives as described (18). The enzyme concentrations were as indicated, and the PAI-1 concentration was minimally 5–10-fold lower to achieve pseudo- first order reaction conditions. Reaction traces were fit by a monoexpo-nential process with a sloping end point to obtain the observed pseudo- first order rate constant ( k obs ) (41). For slower reactions fluorescence kinetic measurements were performed on an SLM 8000 fluorimeter, and data were acquired in the slow kinetic time course mode. Excitation was at 480 nm, and emission was monitored at 529 nm. All experiments were performed in 0.1 M HEPES, 0.1 M NaCl, 1 m M EDTA, 0.1% PEG 8000, pH 7.4. Acrylic cuvettes used for the latter experiments were pre-coated with PEG 20,000 to minimize adsorption. Proteinase concentrations were as indicated, and the PAI-1 concentration was 0.1 or 0.2 m M . In all cases, progress curves were fit well by an exponential process after a brief lag period in some cases. A sloping end point was used to correct for an instability of the fluorescence signal over the longer reaction times required to establish an end point. Kinetic pa- rameters for the interaction of P 9 NBD-PAI-1 with each enzyme were determined by nonlinear regression fitting of the dependence of k obs on proteinase concentration by the hyperbolic Equation 1.

The serpins are a gene family that encode a wide variety of proteins, including many of the proteinase inhibitors in plasma, as well as other non-inhibitory proteins such as steroid binding globulins and ovalbumin (1). The serpins are thought to share a common tertiary structure (2) and to have evolved from a common ancestor (3). Current understanding of serpin structure is based largely on seminal x-ray crystallographic studies by Loebermann and co-workers (4) on one member of the family, ␣ 1 -antitrypsin, also called ␣ 1 -proteinase inhibitor. An interesting feature of this structure was that the two residues normally comprising the reactive center (Met-Ser) peptide bond were found on opposite ends of the molecule, separated by almost 70 Å. Based on this structure, these authors proposed a model structure for native serpins, where the reactive center is part of an exposed loop, referred to as the reactive center loop (RCL) 1 (4). Upon cleavage, the RCL integrates into the center of the dominant serpin structural feature, ␤-sheet A, becoming strand four of this six-stranded ␤-sheet. This transformation is accompanied by a large increase in thermal stability, presumably due to reorganization of the five-stranded ␤-sheet A to a six-stranded anti-parallel form (5)(6)(7)(8). More recent crystallographic structures of several native serpins with intact reactive center loops have confirmed this hypothesis, demonstrating that the overall native serpin structure is very similar to cleaved ␣ 1 -antitrypsin, but that the reactive center loop is exposed above the plane of the molecule (9 -13).
The structural versatility of serpins appears to be essential for their inhibitory activity. Serpins are thought to act as "suicide inhibitors" that react only once with a target proteinase, forming an SDS-stable complex. A "bait" amino acid sequence in the RCL is thought to mimic the normal enzyme substrate and to associate with the specificity crevice of the enzyme (1,14). The amino acid residues toward the NH 2terminal side of the scissile reactive center bond are labeled in order P 1 , P 2 , P 3 , etc., and the amino acids on the carboxyl side of this bond are labeled P 1Ј , P 2Ј , etc. (15). Current models of the serpin inhibitory mechanism suggest that active serpins have flexible RCLs, and that this flexibility is essential for inhibitor function (8, 16 -21). For example, serpin inhibitory activity can be blocked by synthetic peptides that are homologous to the serpin RCL. These peptides incorporate into the serpin, as a central strand of ␤-sheet A (22,23). This leads to an increase in thermal stability similar to that observed following cleavage of a serpin RCL, and converts the serpin to a substrate for its target proteinase (24 -27). While the exact structure of the complex between serpins and their target proteinases is not known, recent data confirm earlier suggestions that the complex is covalently linked via an ester bond between the active site serine residue of the proteinase and the new carboxylterminal end of the P 1 residue, forming an acyl-enzyme complex (18, 28 -32). Moreover, comparison of the substrate reactions of RCL peptide-complexed serpins with the inhibitory reactions of the native serpin have demonstrated that the acylation step limits the rate of both substrate and inhibitor pathways, implying a common pathway for the two reactions up to the acyl-intermediate (26,32). Together, these data have led to a proposed serpin mechanism where proteinase cleavage of the RCL to the point of an acyl-enzyme intermediate is coupled with a conformational rearrangement of the serpin involving RCL insertion. This causes a distortion of the enzyme's catalytic center and consequent trapping of the proteinase (17,18,20,21,30,(32)(33)(34).
In the present study, we use plasminogen activator inhibitor-1 (PAI-1) as a model to test the hypothesis that insertion of the serpin RCL into ␤-sheet A is coupled to proteinase cleavage of the RCL, and that as a result of this coupling serpin inhibition follows a branched pathway leading to either a stable inhibited complex or turnover of the serpin as a substrate. As a critical test of this hypothesis, we have examined whether the partitioning of inhibitory versus substrate activity is governed by the relative rates of RCL insertion versus deacylation of the intermediate acyl-enzyme complex. This was accomplished by perturbing the rates of loop insertion and enzyme deacylation through mutation of the P 14 hinge residue of the RCL, which initiates loop insertion, and through varying the target proteinase. The results presented strongly support the branched pathway mechanism of serpin inhibition and are not consistent with the alternative model of serpin behavior, which suggests that serpins exist in distinct inhibitory and substrate conformations that react by independent pathways (35,36).

EXPERIMENTAL PROCEDURES
Materials-The PAI-1 P 14 mutant library has been described previously (8). All of the single P 14 mutant constructs were expressed, and soluble PAI-1 containing lysates isolated as described (37). The concentration of PAI-1 in Escherichia coli lysates was determined by enzymelinked immunosorbent assay (38). PAI-1 mutants containing a Cys substitution at the P 9 position (Ser-338) of the RCL either in the wild-type PAI-1 background or in combination with substitutions of the P 14 Thr residue (Thr-333) with either Arg, Lys Asp, Glu or His were constructed by site directed mutagenesis as described (39). All of the double P 14 -P 9 Cys mutants were prepared essentially as described (40,41), except that mutants with high stoichiometry of inhibition (SI) values were isolated based upon the usual elution pattern observed with the active mutants, and assessed by SDS-PAGE of individual column fractions during purification. High molecular weight uPA (Ukidan, Serono AB of Switzerland) was a generous gift from Dr. Tor Ny (Department of Medical Biochemistry and Biophysics, Umeå University, Umeå, Sweden), and two-chain tPA was prepared from Activase (Genentech) as described previously (41). ␣-Thrombin was a generous gift of Dr. John Fenton of the New York State Department of Health, Albany, NY, and heparin was from Sigma (grade II from porcine intestinal mucosa).
Inhibitory Activity toward uPA, tPA, and Thrombin-Inhibitory activity toward uPA and tPA was determined as described (33), and the activity toward thrombin was determined using 50 nM enzyme and varying molar ratios of wild-type or variant PAI-1 (up to 40 mol of serpin/mol of enzyme) in the presence of an optimal heparin concentration of 10 g/ml. After incubation for 3 h, residual thrombin activity was determined by chromogenic substrate assay essentially as described (25). The SI toward each enzyme for each of the PAI-1 mutants was calculated from the ratio of the concentration of PAI-1 antigen in each lysate determined in the enzyme-linked immunosorbent assay over the functional concentration of each mutant against each enzyme as determined in the chromogenic inhibition assay (42). The determination of the SI ratio of wild-type PAI-1 over P 14 His PAI-1 as a function of pH was performed as described (8) except that tPA was used as enzyme.
SDS-PAGE and Immunoblot Analysis of Wild-type and Mutant PAI-1s-To determine the extent to which each PAI-1 formed a complex with or was cleaved as a substrate by each enzyme, immunoblot analysis was performed as described (8). Briefly, PAI-1 containing E. coli lysates were diluted to 1 M PAI-1 in 50 mM Tris, pH 7.8, 100 mM NaCl, in the presence of either no enzyme or 1 M uPA, tPA, or thrombinheparin. The samples were incubated for 30 min at 23°C, then subjected to nonreducing SDS-PAGE on a 10% gel (Novex, San Diego, CA), followed by transfer to nitrocellulose and immunoblot analysis developed with a polyclonal rabbit anti-PAI-1.
Fluorescent Labeling of PAI-1 P 9 -Cys Mutants-The purified P 9 Cys mutants were labeled with the environmentally sensitive probe N-((2-(iodoacetoxy)-ethyl)-N-methyl)-amino-7-nitrobenz-2-oxal,3-diazole (NBD) (Molecular Probes) as described (41). Briefly, concentrated samples of each P 9 Cys mutant in 0.05 M sodium phosphate, 0.15 M NaCl, 1 mM EDTA, 0.5 mM DTT, pH 6.6, were applied to a PD-10 gel filtration column (Bio-Rad) equilibrated in the same degassed buffer without DTT. The 280 nm absorbance was monitored, and the PAI-1 was pooled free of traces of contaminating DTT. A 5-10-fold molar excess of NBD was added to the PAI-1 sample and gently mixed. The sample was then incubated at 25°C in the dark for 4 -6 h, after which it was applied to a G-25 superfine column to remove the excess NBD. The labeled PAI-1 was then concentrated by affinity chromatography on heparin-Sepharose and subsequently eluted in 0.05 M sodium phosphate, 1.0 M NaCl, pH 6.6. Quantitation of labeling was as described (41), and the stoichiometry of labeling was between 0.9 and 1.3 mol of NBD/mol of PAI-1 mutant.
Kinetic Analysis of the P 9 -NBD-PAI-1 Variants-For fast reactions stopped flow fluorometry was performed on PAI-1 NBD derivatives as described (18). The enzyme concentrations were as indicated, and the PAI-1 concentration was minimally 5-10-fold lower to achieve pseudofirst order reaction conditions. Reaction traces were fit by a monoexponential process with a sloping end point to obtain the observed pseudofirst order rate constant (k obs ) (41). For slower reactions fluorescence kinetic measurements were performed on an SLM 8000 fluorimeter, and data were acquired in the slow kinetic time course mode. Excitation was at 480 nm, and emission was monitored at 529 nm. All experiments were performed in 0.1 M HEPES, 0.1 M NaCl, 1 mM EDTA, 0.1% PEG 8000, pH 7.4. Acrylic cuvettes used for the latter experiments were pre-coated with PEG 20,000 to minimize adsorption. Proteinase concentrations were as indicated, and the PAI-1 concentration was 0.1 or 0.2 M. In all cases, progress curves were fit well by an exponential process after a brief lag period in some cases. A sloping end point was used to correct for an instability of the fluorescence signal over the longer reaction times required to establish an end point. Kinetic parameters for the interaction of P 9 NBD-PAI-1 with each enzyme were determined by nonlinear regression fitting of the dependence of k obs on proteinase concentration by the hyperbolic Equation 1.
[E] o is the enzyme concentration, and K M is the enzyme concentration required to half-saturate the inhibitor. The constant k lim represents the apparent rate constant for RCL insertion into ␤-sheet A once the PAI-1-proteinase Michaelis complex has formed. For Scheme 1 below, k lim is given by Equation 2 (32).
k a and k Ϫa are forward and reverse rate constants for enzyme acylation, k i is the rate constant for RCL insertion, and k d is the rate constant for enzyme deacylation. When the acylation step is rate-limiting, i.e. k a ϩ k Ϫa Ͻ Ͻ k i ϩ k d , k lim approximates to k a . In the case of the reaction of PAI-1 with thrombin in the presence of heparin, k obs will be dependent on the extent of saturation of both the enzyme and PAI-1 with heparin.
Since the K d for the interaction of thrombin with heparin is 100 -200 nM (43), and since the molar concentration of heparin is ϳ0.7 M, then at the highest concentrations of thrombin used the heparin should be saturated, generating 0.7 M thrombin-heparin binary complex. Since the K d for PAI-1 binding to heparin has been estimated to be ϳ0.5 M, 2 then the fractional saturation of ternary complex will minimally be 0.7 M/(0.5 M ϩ 0.7 M) or 58%. This implies that the observed limiting rate for the interaction of PAI-1 with thrombin-heparin may be underestimated by as much as 2-fold. Although additional kinetic experiments would be necessary to verify the exact extent of this deviation, the calculations nonetheless support the conclusion that the limiting rate for this reaction differs significantly from that of the other enzyme reactions. For analysis of the effects of pH on the rate of RCL insertion with wild-type P 14 -P 9 Cys PAI-1 or with the P 14 His-P 9 Cys PAI-1, 0.1 M PAI-1 and 1 M tPA were used. The experiments were performed as described above for other NBD-labeled variants except that the pH of the reaction was adjusted to the values indicated. The ratio of the pseudo-first order rate constants for loop insertion of P 14 His PAI-1 over that of wild-type PAI-1 were then calculated.

The Conversion of PAI-1 to a Substrate by Mutations at P 14 Is
Enzyme-dependent-Previously we demonstrated that replacement of the wild-type P 14 Thr residue in PAI-1 with a charged amino acid resulted in loss of inhibitory activity toward uPA due to conversion of PAI-1 to a substrate (8). Replacement of the P 14 RCL residue in other serpins with charged amino acids have similarly shown a loss of inhibitor function due to transformation to a substrate (44,45). These findings in conjunction with x-ray crystallographic evidence (4) have suggested that a conformational change involving insertion of the serpin P 14 residue into ␤-sheet A is an obligate step in the formation of a stable serpin-enzyme complex. Furthermore, our observation that the association of a catalytically inactive proteinase with PAI-1 was not affected when the P 14 residue was mutated to Arg, has indicated that the triggering of this RCL conformational change requires the action of the proteinase catalytic residues (46). Together, these and other studies (18,32,(47)(48)(49)(50) have suggested the hypothesis that serpins inhibit proteinases by the branched pathway mechanism outlined in Scheme 1.
According to this mechanism, a proteolytic enzyme, E, first binds reversibly to the RCL of the serpin I, to form a Michaelislike encounter complex, E⅐I. This complex can either dissociate regenerating free enzyme and active inhibitor, or peptide bond cleavage can be initiated, with formation of a covalent acylenzyme intermediate, EI c . This is similar to a proteinase reaction with a regular substrate except that the intermediate may be formed reversibly (32). In this complex, I c is cleaved inhibitor with its RCL still exposed, and covalently tethered to the active site serine of the enzyme. In the ensuing steps in this pathway, there are two potential outcomes for the EI c complex. In the first, the serpin RCL rapidly inserts into ␤-sheet A generating the EI i complex, wherein the inhibitor is cleaved, and the RCL is fully inserted, with the inhibitor still tethered to the enzyme as a covalent acyl-enzyme intermediate. This results in stabilization and trapping of the covalent complex presumably due to distortion of the enzyme's active site (34,51). Alternatively, the EI c complex can undergo deacylation with release of the active enzyme before serpin RCL insertion. The free cleaved inhibitor with an uninserted RCL (I c ) can then insert into ␤-sheet A in a non-productive conformational rearrangement that generates an irreversibly inactivated serpin (I i ).
This hypothesized model suggests that it is the competing rates of RCL insertion (k i ) versus enzyme deacylation (k d ) that govern the relative partitioning between the two potential outcomes. To test this hypothesis, we examined the partitioning between formation of a stable EI i complex and turnover of the serpin as a substrate by determining the SI for a library of PAI-1 P 14 mutants with three different enzymes. The SI is defined as the moles of inhibitor required to inhibit 1 mole of enzyme. An SI of 1 indicates that 100% of the inhibition reaction was productive, with all of the serpin partitioning to the EI i complex, whereas an SI of 10 indicates that 90% of the inhibitor was turned over as a substrate, and only one serpin molecule in ten formed a stable EI i complex. If this mechanism is correct, then the SI of a serpin proteinase reaction should be determined by the ratio of k d to k i (SI ϭ 1ϩ k d /k i ). Thus, variation in either rate constant should alter the SI of the reaction as long as this variation results in k d approaching or exceeding k i .
We chose to vary k i by taking advantage of a previously described library of PAI-1 P 14 point mutants (8). This library contains all potential amino acid substitutions at the P 14 position except for Pro, Trp, and Ser. Based on our earlier analysis, we postulated that P 14 residues with different hydophobicities would insert into the hydrophobic core of PAI-1 at distinct rates and thus result in altered rates of RCL insertion (k i ). To deter-mine whether different rates of deacylation (k d ) would also affect SI, we chose to examine this library of mutants with three different proteinases, uPA, tPA, and thrombin, based on the expectation that these proteinases would show dissimilar rates of deacylation. Fig. 1 shows the SI values measured for the reactions of the P 14 PAI-1 mutants with the three target proteinases. As predicted from the branched pathway model, the SI varies greatly either when PAI-1 molecules with different P 14 residues react with the same proteinase, or when different proteinases are inhibited by the same PAI-1 mutant. For example, uPA (open bars) has a low SI value with all P 14 mutants except those with charged residue substitutions (Arg, Lys, Asp, Glu), whereas both tPA (black bars) and thrombin (hatched bars) have SI values that are 10 -100-fold higher than uPA with all of the neutral polar residue substitutions (His, Tyr, Gln, Asn) or with a Gly substitution which lacks a side chain. In contrast, when the P 14 residue is hydrophobic (Val, Met, Leu, Ile, Phe), none of the proteinases show a large difference in SI compared with wild-type PAI-1. The three proteinases also behave similarly in showing uniformly high SI values when the P 14 residue is charged. These results are consistent with the expectation that hydrophobic P 14 residues, which should favorably insert into the hydrophobic core of the protein, are associated with low SI values, whereas neutral or charged polar residues, which should not insert as readily into the hydrophobic core, or Gly, which has no side chain to insert, are associated with higher SI values that are enzyme-dependent. The slight decrease in SI for some of the mutants compared with wild-type PAI-1 with each of the enzymes most likely reflects a reduced tendency for these PAI-1 variants to adopt the latent conformation, since the transition from active to latent PAI-1 also requires insertion of the RCL (13,52,53). Finally, the somewhat higher SI values seen with thrombin-heparin and wild-type PAI-1 are very similar to previously reported results, demonstrating that heparin increases the SI of both PAI-1 and antithrombin III reactions with thrombin. This effect also appears to arise from a reduced rate of RCL insertion due to the need for thrombin to dissociate from heparin to allow the RCL and attached enzyme to insert into sheet A (54, 55).
The dependence of SI on the choice of enzyme as seen with all of the neutral polar residue substitutions is completely consist-

FIG. 1. Effect of P 14 residue mutations on the stoichiometry of PAI-1 inhibition of proteinases is both enzyme-and mutant-dependent.
The SI values for each of the P 14 mutants are shown versus uPA in the white bars, tPA in the black bars, and thrombin-heparin in the hatched bars. The SI was calculated as described under "Experimental Procedures," and the error bars represent the S.E. Bars without errors represent the limit of sensitivity obtained with each enzyme and, thus, the minimum SI value for a given mutant with that enzyme. The maximum SI that could be determined for uPA was 600, that for tPA was 200, and that for thrombin-heparin was 800. ent with the branched pathway model of serpin function, and entirely inconsistent with a model of serpin behavior where substrate and inhibitory activities are predetermined by the conformation of the inhibitor (35,36). This is because in a branched pathway, different enzymes with differing rates of deacylation can alter the ratio between substrate and inhibitory activities as discussed above. However, if serpin substrate behavior results from the conformation of the inhibitor, then simply changing enzymes should not alter the relative amounts of inhibitory and substrate conformations and thus the SI should not be enzyme-dependent. Fig. 2 shows an immunoblot of the products formed in the reaction of the PAI-1 P 14 mutants with each proteinase after separation by SDS-PAGE. The majority of PAI-1 can be seen in each case to be either cleaved as a substrate or to form an SDS-stable complex with the proteinase, indicating that there is relatively little nonreactive latent PAI-1 in any of the mutant preparations. In addition, even though the immunoblots are only semiquantitative, the observed extent of complex formation is also in approximate agreement with the calculated SI values shown in Fig. 1. These results indicate that the reason for the different SI values obtained for the mutant PAI-1 proteinase reactions is primarily due to partitioning between enzyme inhibition and enzyme turn-over, rather than to varying amounts of nonreactive or latent PAI-1 in each preparation (52). Together, the data of Figs. 1 and 2 support the idea that the PAI-1 P 14 residue mutations affect SI by altering the rates of RCL insertion into ␤-sheet A, and that the extent to which the SI is affected depends on the different efficiencies of enzyme deacylation. Such data thus fulfill the predictions of our hypothesized mechanism in which the relative rates of RCL insertion and enzyme deacylation govern the partitioning between inhibitor and substrate modes of serpin reaction.
The Rate of RCL Insertion Is Proteinase-dependent-We previously showed that a P 9 Ser 3 Cys variant PAI-1 could be specifically labeled in the RCL with the environmentally sensitive fluorophore, NBD, and that this label reports the RCL conformational change through a large enhancement and blueshift in the fluorescence emission spectrum (18,41). The fluorescence changes presumably reflect burying of the solventexposed P 9 side chain beneath ␣-helix F and the adjoining extended loop when the RCL inserts into sheet A. This assumption is based on the structural changes seen to accompany such insertion in comparing native with cleaved and latent forms of PAI-1 (13), and is consistent with our previous observation that the fluorescence changes induced in the labeled PAI-1 by RCL cleavage or conversion to the latent form are similar to those induced when the labeled PAI-1 is complexed with proteinases (41). Finally, fluorophore labeling did not perturb PAI-1 func-tion since the purified, NBD-labeled mutant had nearly wildtype inhibitory activity against uPA and tPA with second-order inhibition rate constants of 7 ϫ 10 6 M Ϫ1 s Ϫ1 and 14 ϫ 10 6 M Ϫ1 s Ϫ1 , respectively (33). Fig. 3 shows the observed pseudo-first order rate constant (k obs ) for the RCL conformational change induced in labeled PAI-1 during its reaction with proteinase, as measured by continuous monitoring of NBD fluorescence changes, plotted as a function of proteinase concentration with uPA, tPA, thrombin-heparin, or thrombin alone. The k obs increases in a saturable manner with increasing proteinase concentration to a limiting value in all cases, consistent with the RCL conformational change being induced subsequent to the formation of a reversible serpin-proteinase encounter complex. The limiting rate constant (k lim ) represents the apparent rate constant for RCL insertion into ␤-sheet A once the PAI-1-proteinase Michaelis complex has formed. The k lim is expected to be a function of the acylation, RCL insertion, and deacylation rate constants for the mechanism of Scheme I (32). It is notable that the limiting rate constant is greatly dependent on the enzyme, with uPA showing the fastest rate and thrombin the slowest. The kinetic parameters for the interaction of P 9 NBD-PAI-1 with each enzyme were determined by nonlinear regression fitting of the data in Fig. 3 by Equation 1 (see "Experimental Procedures"). These parameters are shown in Table I, and indicate that k lim for the different enzymes vary by more than 2 orders of magnitude, ranging from 0.036 s Ϫ1 to 5.9 s Ϫ1 . These results are similar to those previously reported comparing the limiting rates of PAI-1 acylation by tPA and trypsin (32). The data are also consistent with the hypothesized mechanism of Scheme I and with our previous findings that insertion of the RCL into ␤-sheet A requires prior RCL cleavage by an active proteinase and is limited by the rate of enzyme acylation (18,26,32,46). The different rates of RCL insertion induced by each enzyme are thus readily explained if the rate of RCL insertion is limited by the rate of RCL acylation, since the acylation rate is expected to be unique for each enzyme. That the kinetic parameters shown in Table I accurately reflect the interaction of wild-type PAI-1 with each enzyme is suggested by the agreement between apparent second-order rate constants (k lim /K M ) calculated for the interactions of the labeled PAI-1 with each enzyme and those reported for inhibition of each enzyme by wild-type PAI-1 (33,56).
Mutations at P 14 Affect the Rate of RCL Insertion-To determine whether mutations of the PAI-1 P 14 residue directly affect the rate of RCL insertion induced by a proteinase, and thereby promote the substrate reaction of the mutant PAI-1s, we constructed additional PAI-1 mutants containing both a P 9 Cys substitution together with a P 14 Arg, Lys, Glu, Asp, or His replacement. These constructs were purified, labeled with NBD and reacted with tPA, uPA and thrombin/heparin. Fig. 4 shows a comparison of the results obtained for the reactions of "wild-type" P 9 NBD-PAI-1 (panel A), and a P 14 Arg-P 9 NBD-PAI-1 double mutant (panel B) with tPA. These data demonstrate a very similar increase in fluorescence for both P 9 NBD-PAI-1 species, suggesting that, like wild-type PAI-1, the P 14 Arg mutant can also insert its RCL into ␤-sheet A. Furthermore, since the magnitude of the fluorescence change is similar to wild-type PAI-1, the NBD reporter group appears to be in a comparable environment in both cases. This suggests that the RCL of the P 14 Arg mutant is most likely inserted into ␤-sheet A to a similar extent as wildtype PAI-1, and is consistent with the crystal structure of a RCL-cleaved P 14 Arg mutant of ␣ 1 -antichymotrypsin (57). In this structure the RCL was inserted into ␤-sheet A, although the P 14 Arg residue was distorted from its usual position with the sidechain atoms exposed to the solvent. However, while the extent of RCL insertion appears similar when tPA is reacted with P 9 NBD-labeled wild-type and P 14 Arg PAI-1, comparison of panels A and B of Fig. 4 indicates that the rate of insertion is dramatically slower when the P 14 residue is replaced with Arg, increasing the t 1/2 for insertion nearly 3500-fold from ϳ0.3 s to over 15 min. This suggests that the conversion of P 14 Arg PAI-1 to a substrate results from the reduced rate of RCL insertion following the initial interaction with a target protease.
Additional analyses of the reactions of the other NBD-labeled double P 14 -P 9 mutants with uPA, tPA, and thrombinheparin support this conclusion. Table II reports k obs values for RCL insertion during the reactions of the labeled PAI-1 P 14 variants with different proteinases at 0.1 and 1 M proteinase. These results demonstrate that for all four PAI-1 variants with charged substitutions at P 14 the observed rate of RCL insertion is essentially independent of proteinase concentration and is greatly reduced compared with wild-type PAI-1. Moreover, the RCL insertion rates for any one mutant PAI-1 are similar for different proteinases. This contrasts with the marked dependence of the rates of RCL insertion for the P 9 NBD-labeled wildtype PAI-1 on the specific proteinase inhibited and the proteinase concentration (Table II and Fig. 3). Such findings suggest that the k obs values for the charged variant PAI-1s represent the intrinsic slow rate of RCL insertion after the serpin RCL scissile bond has been rapidly cleaved by proteinase as a regular substrate. Together, these results suggest that when the intrinsic rate of RCL insertion is faster than the rate of acylation, as is the case with wild-type PAI-1, then the observed rate of RCL insertion will be limited by the rate of acylation and thus be dependent on both the proteinase and its concentration. However, when the rate of acylation is faster than the intrinsic rate of RCL insertion, as with the charged P 14 mutants, then the observed rate of RCL insertion is the true, or intrinsic rate, and therefore no longer dependent on either the proteinase or its concentration.
The Rate of RCL Insertion and the SI Are Linked-In previous studies we demonstrated that reducing the pH from 9.0 to 5.0 markedly decreased the inhibitory activity of a P 14 His mutant toward uPA, while similar treatment of wild-type FIG. 3. Stopped flow kinetic analysis of P 9 -NBD PAI-1 reactions with uPA, tPA, or thrombin. Panel A, pseudo-first order rate constants (k obs ) for the reactions of 0.005-0.1 M P 9 NBD PAI-1 with uPA (Ⅺ), tPA (E), thrombin (‚), and thrombin-heparin (ƒ) are plotted as a function of the enzyme concentration. Panel B shows the curves for thrombin and thrombin-heparin on an amplified scale. Solid lines are fits by the hyperbolic equation given in the text. It should be noted that at the optimal heparin concentration employed for the reaction (0.7 M), k lim may reflect saturation of the binary thrombin-heparin complex without complete saturation of the ternary complex. However, given reasonable estimates of the PAI-1-heparin affinity, k lim should be underestimated by no more than 2-fold (see "Experimental Procedures").
FIG. 4. Kinetic analysis of NBD fluorescence changes for wildtype and P 14 -Arg PAI-1 reactions with tPA. Panel A shows P 9 NBD PAI-1 wild-type P 14 (q), and panel B shows the P 9 NBD PAI-1 P 14 Arg mutant (q). In both cases 0.1 M labeled PAI-1 was reacted with 2 M tPA. Solid lines are fits by an exponential function with a sloping end point.

FIG. 5.
Comparison of the effect of pH on the ratio of rate constants for RCL insertion and the ratio of SIs for wild-type and P 14 -His PAI-1 reacting with tPA. Both the ratio of the rate constants (E), and the ratio of the SIs (Ⅺ), determined as described under "Experimental Procedures," were fit for a single pK a (8). The fit for the rates is shown with the dashed line, and the fit for the SIs is shown with the solid line. a Due to the incomplete saturation of the ternary thrombin-heparin-PAI-1 complex the observed k lim for the interaction of PAI-1 with thrombin-heparin is likely to be underestimated by as much as 2-fold (see "Experimental Procedures"). PAI-1 had a much smaller effect on its activity (8). Since the His residue is structurally the same at pH 9 and pH 5 except for its protonation state, these data suggest that charge was the critical factor in determining whether a mutant behaved as an inhibitor of uPA or as a substrate, and that as the His residue became charged its rate of RCL insertion was reduced and therefore its SI increased. To test this possibility directly, the ratio of the rate constants for RCL insertion for a P 14 His-P 9 NBD-PAI-1 to that for "wild-type" NBD-PAI-1 was calculated for the pH range 5.5-9.0 and compared with the ratio calculated for the SI values of wild-type PAI-1 over a P 14 His PAI-1 mutant for the same pH range, using tPA as the enzyme.
Calculating the ratio served to cancel out potential effects of pH on the inhibitory activity due to protonation of the proteinase catalytic histidine 57 or of any other histidine residues within the enzyme or the inhibitor. These data are shown in Fig. 5 and demonstrate that both ratios change in a parallel manner over the pH range examined. Furthermore, the pK a values for these data were evaluated and yielded values of 6.0 and 6.1 for the ratios of SI and rate, respectively. The parallel tracking of these two properties from pH 9.0 to 5.5 strongly suggests that they are related such that as the rate of RCL insertion decreases, the value of SI increases relative to wild-type PAI-1. The simplest explanation for these data is that the charging of the P 14 His as the pH is lowered changes the SI by decreasing the rate of burial of the P 14 side chain into ␤-sheet A in a manner similar to that observed for the charged P 14 residue variants of PAI-1. Taken together, these data strongly support the hypothesis that serpins inhibit proteinases by a branched pathway mechanism, where the partitioning between trapping a proteinase in a stable inhibited complex or turnover of the serpin as a substrate can be altered by perturbing either the rate of RCL insertion into ␤-sheet A, through changes in the P 14 residue, or the rate of deacylation, through changes in the proteinase. These results provide the first direct evidence that inhibition and substrate modes of serpin reaction with proteinases arise from competing rates of RCL insertion versus enzyme deacylation of a common acyl-enzyme intermediate in accordance with the branched pathway mechanism of Scheme I, and not from a unique serpin substrate conformation that has no intrinsic inhibitory activity (35,36). Our results further suggest that the essential features of the branched pathway mechanism include (i) RCL cleavage by a proteinase inducing RCL insertion into ␤-sheet A and (ii) RCL insertion being absolutely required for stable inhibition. This mechanism is also consistent with previous observations that serpin SIs are temperature-dependent, showing higher SI values at lower temperature (47,58), as would be expected if the conformational change associated with RCL insertion were much more temperature-sensitive than the chemical step of deacylation. Thus, these findings are likely to provide a general model for serpin inhibitory function.